LX-mini Analog Crossover - firstwatt.com

1 LX-mini Analog CrossoverMany of you are already familiar with Siegfried Linkwitz. I first met him in personat a Burning Amp Festival years ago, and I was able to spend more time with him after I moved to Sea Ranch, where he has a house at the far end of the beach. We have spent many afternoons on his deck in the sun, talking about audio and snacking on the treats provided by his wife Eike. Among his numerous loudspeaker designs is the LX-mini , and there is a demo pair installed in his living room. It's a remarkable loudspeaker, biamped with a DSP based Crossover from MiniDSP with filter parameters designated by Siegfried.

2 It would be difficult to make it work with a passive will not recapitulate the design of the Lxmini here, rather I will send you to his web site and another site where there are complete discussions about his philosophy and loudspeaker fine day over melon and cookies he asked if I might have an interest in working up an Analog version of the Crossover , as he had some interest from hisusers. I took a look at the DSP filter curves, and said I dunno, Siegfried, these are some pretty complicated curves you got here.

3 It's gonna take a lot of op amps and stuff, and you know me I only do the simple stuff. Persevering, he later offered up a copy of the 2011 edition of Douglas Self's The Design of Active Crossovers , offering 570+ pages of late night reading, and more than you might want to know about Crossover filters. To appreciate the difficulty of replicating his DSP curve, you might want to look at it:It's got lots of little bumps, peaks and dips, each correcting for the curve that Siegfried obtained when he put a microphone in front of the speaker and measured the response.

4 I'm not going to chase that stuff, I said, but after a couple of weeks I decided that it would be interesting to do something with broad strokes and see how it worked out. After some time with Micro-Cap I had a decent looking simulation, so I laid out a circuit board with Jfet buffers and gave Siegfried a working copy to play with. He came back with some suggestions for curve modifications based on his listening, and later measured the response in comparison with the DSP based is the circuit we ended up with, as seen by the simulator.

5 X1 through X4 are op amp based buffers representing the simple Jfet buffers. At the top we see the equalization and then the high pass filter for the mid/high driver, and below that is the low pass filter and then bass equalization for the low/mid , the filter topology is remarkably close the First Watt B5 Crossover , which caters to full range drivers and woofers in open baffle enclosures. This circuit produces a curve that looks like this:Initially SL had some concern that we were not completely duplicating the curves of the original, particularly at the top end, where there are high Q peaks and dips.

6 However, he subsequently measured the acoustic response and found that the top end was a bit flatter without those corrections. Here is that result with the DSP on top and Analog on the speculated a bit on why that might be. One possibility was that the anomalies of the high end driver might vary from unit to unit. The other was thatat wavelengths of less than an inch, variation of microphone position might account for above curve also shows the response of both with the speaker drivers wiredin-phase, where ideally they should have a deep null at the Crossover point, a good indicator that the phases are identical at the frequency in normal reminds me to mention that there are no phase inversion circuits in this Crossover , only unity gain discrete Jfet buffers without feedback.

7 This means that with this Crossover you have to invert the phase of the mid-tweeter, probably in the speaker wiring between the amp and the mid-tweeter, otherwise you will be listening to that big deep the bottom you can note that the Analog Crossover drops off faster below 40 Hz, an artifact of generating the 50 Hz equalization peak with a simple buffer. I didn't expect this to please SL, but he didn't complain when we played the Telarc1812 Overture cannon shots, so I guess it's okay, and the drivers didn't bottom have mentioned the Jfet buffers, and as there are some interesting aspects to the design, it is worth talking about them.

8 As you may know, a buffer has the same voltage at the output as its input voltage, but it increases the amount of current available. Many types of audio filters can be made with buffers, and I particularly like them because they are very simple with high performance. Hereare two favorite Jfet topologies:The buffer on the left has two complementary Jfets in Common Drain (follower) mode. On the positive side is an N channel Jfet sharing the input and output with a P channel Jfet in what is known as a push-pull the right are two N channel parts, with the top one acting as the buffer transistor and the bottom one is set up as a constant current the 1980's there have been favored parts from Toshiba, the 2SK170 and 2SJ74 Jfets, which work very well in such simple circuits, giving line level distortion numbers in the low region without any special treatment.

9 They have been very popular with DIYers, but were unfortunately discontinued afew years ago. You can still get them, but they are scarce and the price has gone up by as much as a factor of 50. There are good replacements from Linear Systems, but these are also expensive and generally in short of Siegfried's goals for the LX-mini was that it would not be very expensive. Many of the parts are available at hardware stores, and the cost for a complete pair is well less than $1,000, including the DSP Crossover . I undertook to find aninexpensive Jfet that could be made to give good bought samples and played with them, ultimately settling on the Fairchild (OnSemi) J113, which is cheap and available.

10 It can be used in the single-ended circuit above, but without part selection / matching and some additional tweaking it will not give particularly good numbers distortion figures from to Matching the devices improves that number, but they still fall short of the Toshiba / Linear Systems numbers, although they are very is a schematic showing a pair of J113 Jfets used as a buffer for crossoverfilters. The two devices have been matched and then a resistor R1 has been used to set the bias current and another resistor to load the output of the buffer at a specific value.